running head: ssi and ssdi usage among americans nearing ......disability theories, and disability...

Running head: SSI and SSDI Usage among Americans Nearing Retirement 1

Work Transitions among SSI and SSDI Recipients Nearing Retirement: Employer

Accommodations, Private Insurance, and Individuals with Mental Health Disability

Mitzi K. Lauderdale, J.D., CFP®

Doctoral Student

Kansas State University

Stuart J. Heckman, Ph.D., CFP®

Assistant Professor and Faculty Mentor

Kansas State University

The research reported herein was performed pursuant to a grant from Policy Research, Inc. as

part of the U.S. Social Security Administration’s (SSA’s) Analyzing Relationships Between

Disability, Rehabilitation and Work. The opinions and conclusions expressed are solely those of

the author(s) and do not represent the opinions or policy of Policy Research, Inc., SSA or any

other agency of the Federal Government.

SSI and SSDI Usage among Americans Nearing Retirement 2

Abstract

Disability diagnoses and US disability program expenditures have been on the rise for several

decades. Despite a leveling off in 2015, a great burden continues to be placed on both the

Supplemental Security Income (SSI) and Social Security Disability Insurance (SSDI) program

funds. Even after the Americans with Disability Act of 1990, employment rates and workplace

accommodations among disabled individuals remain low. Demand for private long-term care

insurance remains low while the need to receive care for the long-term is on the rise. Individuals

with mental illness are the fastest growing segment of SSI and SSDI recipients. Individuals with

mental illness are less likely to receive work accommodations and face greater stigmas due to their

disability. Using Mitra’s (2006) adaptation of the capability approach and panel data from the

Health and Retirement Study (HRS) and psychosocial leave behind survey, this study explores

individuals nearing retirement and their use of SSI and/or SSDI. Three random effects logistic

regressions were employed to predict use of a) SSI and/or SSDI (Model 1), b) SSDI (Model 2),

and c) SSI (Model 3). Private long-term care insurance, employer provided health insurance,

personality traits (Big 5 traits), and psychiatric diagnosis were examined in the regressions. As

Mitra suggests in the capability approach, the choices an individual makes can result in actual

functioning or lack thereof. Those who have a psychiatric diagnosis have significantly higher odds

of using SSI and/or SSDI. Having access to employer or spousal employer healthcare insurance

and being higher on the Big 5 conscientiousness scale reduces the odds of using SSI and/or SSDI

while and increase in the Big 5 agreeableness increases the odds of using SSI and/or SSDI. Lastly,

choosing to purchase private long term care insurance significantly lowers the odds of using SSI.

Policy implications were identified including incentives for individuals and/or employers to carry

or provide private long-term care and health insurance for families, and continuing higher

education incentives which all could result in some relief to the SSI and SSDI program funds.


Background

Introduction

Disability diagnoses affect an estimated 13% of the US population (Erickson, Lee, & von

Schrader, 2018). The financial cost of having a disability varies in impact on the individual,

family, and community depending on one’s employability and severity of disability. Disability

diagnoses and US disability program expenditures were on the rise for several decades. Despite

some leveling out, the Social Security Disability Insurance Trust Fund is fiscally imbalanced and

estimated to be depleted by 2032 (Social Security and Medicare Boards of Trustees, 2018).

Demand for private long term care insurance remains low, yet the demand for long term

care service is rising along with the number of disabled individuals and individuals nearing

retirement. Some research suggests individuals are less likely to have long term care insurance if

they believe there is a substitute for the coverage (Brown, Goda, & McGarry, 2012). Other studies

examined the demand effects of private long term care insurance and found Medicare and

Medicaid to be considered substitutes for private long term care insurance and used it to explain

why wealthy individuals were less likely to purchase private long term care insurance even though

they were the ones who have the means to do so (Brown & Finkelstein, 2008; Pauly, 1990).

Unfortunately, those considering Medicare as a substitute are mistaken as to the coverage

provided.

Personality traits have been studied extensively by psychologists (Caspi, Roberts, &

Shiner, 2005) and applied across many other fields and to predict important life outcomes

(Roberts, Kuncel, Shiner, Caspi, & Goldberg, 2007). The Big 5 personality traits of extroversion,

agreeableness, conscientiousness, openness to experience, and neuroticism are constructs which

have been tested on adults and children, across cultures, and as individuals age (Mroczek, Spiro, &

Griffin, 2006). As it relates to disability, researchers found that one’s specific personality trait of

agreeableness prior to a disability onset can influence their adaptation and recovery of life

satisfaction (Boyce & Wood, 2011).

This study explores SSI and SSDI usage by examining whether having private long term

care insurance, employer provided health insurance from the respondent’s or spouse’s employer,

having received a mental health diagnosis, and variations among personalities (Big 5 traits) predict

usage. Three random effects logistic regressions were employed to predict use of a) SSI and/or

SSDI (Model 1), b) SSDI (Model 2), and c) SSI (Model 3).

Literature Review

US disability program expenditures place a great burden on the Supplemental Security

Income (SSI) and Social Security Disability Insurance (SSDI) program funds. The SSDI Trust

Fund is funded through social security disability insurance payments by workers and employers,

and SSDI qualification is based on entitlement through having paid into the fund. While the total

blind and disabled SSI expenditures began leveling out in 2015, expenditures on behalf of

individuals aged 50 and above have continued to rise (2017 Annual Report of the SSI Program).

SSI recipients qualify based on being blind or disabled or alternatively by reaching age 65. In


addition, an income test and asset test must be met to qualify for the benefits. Since this is a form

of welfare, the benefits put strains on the general funds from the US Treasury to fund

Supplemental Security Income (SSI).

Researchers have found that even though capacity has diminished for many recipients,

disabled individuals often have remaining work capacity and the receipt of disability benefits

could discourage labor force participation (Maestas, Mullen, & Strand, 2013). Mental illnesses

comprise the largest group of those receiving SSI and SSDI, have surpassed growth in other

categories, and many recipients who identify primarily with other types of disabilities are also

facing psychiatric comorbidities which present problems returning to the labor force (McAlpine &

Warner, 2000).

Long-term care insurance policies are available for purchase as private insurance and

Medicaid provides long-term care benefits for SSI participants. Some researchers have suggested

that private long-term care insurance for less affluent individuals has been crowded out by

Medicaid (Sloan & Norton, 1997). Demand for private long-term care insurance in elder law

planning often includes legal spend down plans to qualify for SSI in order to obtain long term care

coverage. Once an individual is qualified for SSI, they are qualified for Medicaid long-term care

benefits. If individuals had quality private long-term care benefits, the need to spend down assets

and qualify for Medicaid would lessen. According to the Genworth Cost of Care Survey (2016),

the median semi-private nursing care costs $6,844 per month and median private room cost $7,698

per month and costs are rising rapidly. Long-term care is usually needed for most types of

disabilities and includes mental illnesses.

Individuals with mental illness are often undiagnosed. However, even if diagnosed, they

often withhold the diagnosis from family, friends, and employers due to strong social stigmas.

Some researchers have even suggested referring to mental illnesses as brain diseases due to the

stigma and discrimination related to mental illnesses (Corrigan & Watson, 2004). Despite

legislative efforts to decrease discrimination of disabled individuals in the workplace, employment

rates of disabled individuals have remained low despite the fact that an overwhelming majority of

unemployed disabled workers say they prefer to work (Unger, 2002). While work does consume a

lot of one’s time, it also plays a vital role in maintenance of mental health, development, and

expression (Blustein, 2008). Working can provide more than economic value to individuals and

has been shown to extend to life satisfaction, especially among those with mental illness (Laughlin

& Cotton, 1994).

The Americans with Disability Act of 1990 (ADA) requires employers to provide

reasonable accommodations to disabled individuals as long as it does not cause an undue burden

to the employer (Stein, 2003). Even after ADA, employment rates of disabled individuals remains

low (Burkhauser & Stapleton, 2004; Solovieva, Dowler, & Walls, 2011). Accommodations in the

workplace to support individuals with disability have been shown to delay the timing of

applications for SSDI (Burkhauser, Butler, Kim & Weathers, 1999) and researchers suggest those

who receive employer accommodations are 40% more likely to postpone their exit from the labor

market for up to two years (Hill, Maestas, & Mullen, 2016).


Furthermore, employers who make accommodations report direct benefits such as

employee retention, reduction in training costs, and increased worker productivity resulting in an

average of $1,000 savings per employee along with indirect benefits such as morale, coworker

interactions, and overall productivity (Solovieva, Dowler, & Walls, 2011). Hill et al. (2016) also

found employee personality types were highly predictive of whether an employer provided

accommodations or not. While only a quarter of employers provide accommodations for disabled

workers over the age of 50, employees may be partially responsible because many do not ask for

accommodations (Hill, Maestas, & Mullen, 2016). Individuals with disability are more likely than

those without disability to be in part-time or contingent jobs and Schur (2003) suggested this is

due to the income limitations placed on individuals receiving SSI and SSDI, employer

discrimination, and actual disability limitations preventing full time employment.

This study investigates individuals nearing retirement and their use of SSI and/or SSDI.

Research questions are focused on how having long term care insurance, access to employer

provided (respondent or spouse) health insurance, having a diagnosed mental illness, and varying

levels or personality traits related to the use of SSI and/or SSDI.

Theoretical Framework

Theoretical approaches have varied among researchers when examining disability, but

most commonly used have been disability models including the medical model, the social model,

and the Nagi Model (Altman, 2001; Mitra, 2006). Others have utilized economic models such as

the life cycle hypothesis (Low & Pistaferri, 2010). Bernell (2003) reviewed disability definitions,

disability theories, and disability laws and provided a four pronged approach to defining disability.

Bernell synthesized definitions and suggested the inclusion of (a) the presence of a physical or

mental condition, (b) receipt of disability benefits, (c) measures of functional ability, including

work, and (d) accommodation and type of accommodations in the workplace.

The capability approach was recommended by Mitra (2006) over medical models because

it is more comprehensive and takes into account a variety of factors including resources available,

the environment and personal characteristics. The capability model, shown in Figure 1, can be

applied to disability and focuses more on capabilities which lead to functioning. The capability

model incorporates all of Bernell’s (2003) recommendations and is much more comprehensive.

Ultimately, Mitra (2006) suggests disability can be analyzed as deprivation of capabilities

(potential disability) or a deprivation of functioning (actual disability). The difference between

capabilities and functioning are one’s choices or adjustments based on values. According to Mitra

(2006), adjustments to abilities are sometimes made in the face of deprivation.

Based on literature reviewed and Mitra’s capability approach, I hypothesize that

individuals without long term care insurance will have higher odds of utilizing SSI. In general,

utilization of government benefits will vary among personality traits. More specifically,

individuals with higher levels of extroversion, agreeableness, conscientiousness, and openness will

have will have lower odds of utilizing SSI and/or SSDI and individuals with higher levels of

neuroticism will have higher odds of utilizing SSI and/or SSDI. Lastly, it is expected that

individuals with health insurance coverage provided by their spouse’s or one’s own employer will

have lower odds of using SSI and/or SSDI.


Research Design, Methods and Data Analysis

The Health and Retirement Study (HRS) is the largest comprehensive, nationally

representative longitudinal panel survey of Americans over the age of 50. Panel interviews were

conducted every other year of the first cohort since 1992. Other cohorts were added totaling seven

cohorts, and a continued plan to add new younger cohorts every six years. Respondents provided

detailed information on health, psychosocial information, employment history, financial

information, demographic information, and benefits details for SSI and SSDI on the respondent

and spouse which makes the data ideal for this study. The RAND Center for the Study of Aging,

with funding from the National Institute on Aging and the Social Security Administration,

generates a cleaned and user friendly version of the HRS. The RAND data were merged with the

psychosocial leave behind files and core content related to work accommodations found in the

core M1 files.

Some respondents authorized administrative data links to the Social Security

Administration (SSA) which was utilized as a follow up to compare the analysis using the self-

reported data in the publicly available set to the SSA linked data for SSI. In phase two of the

project, the publicly available data was utilized and later combined with the restricted data for

further in depth analysis into SSI usage. Literature suggests there may be a selection issue with

someone opting into the linked data by providing permission. Those with administrative links

have been found to have greater education and higher levels of income than those without links

(Kapteyn, Michaud, Smith, and Van Soest, 2006). If only the linked data were used, this would

have needed to be taken into account and addressed by a selection model. However, the linked

data were combined with the publicly available data and simply used to clarify some missing data

related to the dependent variables. The HRS data are publicly available, but the SSA linked data

requires a restricted data application (RDA) for a license. IRB and MiCDA enclave approvals

were obtained for the second phase of the research where the restricted data were incorporated.

The data used in the following five models come from five different waves collected every

other year from 2006-2014. The full sample during this decade includes approximately 20,000

respondents. The sample in these models were restricted to respondents under the age of 65

reducing the sample to approximately 7,000 individuals between the age of 50 and 65. Under 65

was the population of interest because SSI becomes available to individuals without disability

after the age of 65 based on the income and asset requirements and SSDI is only available until an

individual reaches full retirement age which varies among respondents but all exceed 65 years.

The psychosocial leave behind questionnaires were only administered to 50% of the sample each

wave, conducted every other year, and were combined into time invariant variables derived out of

years 2008 and 2010.

Three separate models were employed with different dependent variables with consistent

variables of interest and control variables with one fewer variable of interest in Model 3. The

capability approach was operationalized using commodities/resources of household income,

household net worth, health insurance, and long term care benefits. Environment factors

addressing physical, social, economic, cultural, political included race and census region. Personal

characteristics included variables for psychiatric diagnosis, health status, gender, education,


marital status, and Big 5 personality traits. Capabilities were measured by time variant labor forces

status. Lastly, functioning was measured by the use of SSI and/or SSDI benefits.

Dependent Variables

Model 1: Use of SSI and/or SSDI. The binary variable was coded as 1 for each wave if the

respondent reported they were applying for SSI, receiving SSI, applying for SSDI, applying for

both, applying for SSDI/receiving SSI, receiving SSDI, receiving SSDI/applying for SSI,

receiving SSDI and SSI, applying but don’t know which, or receiving but don’t know which. One

of the biggest differences in the dependent variable for Model 1 when compared to Models 2 and 3

are two response options of applying but don’t know which (SSI or SSDI) and receiving but don’t

know which (SSI or SSDI). Application and receipt of benefits were both included because they

are both an intentional choice to use the benefits.

Model 2: Use of SSI. The binary variable was coded as 1 for each wave if the respondent

reported they were applying for SSI, receiving SSI, applying for both, applying for SSDI/receiving

SSI, receiving SSDI/applying for SSI, and receiving SSDI and SSI.

Model 3: Use of SSDI. The binary variable was coded as 1 for each wave if the respondent

reported they were applying for SSDI, applying for both, applying for SSDI/receiving SSI,

receiving SSDI, receiving SSDI/applying for SSI, and receiving SSDI and SSI.

The n varies across waves by decreasing due to panel attrition and increasing due to

additions of new cohorts. As show in Table 1, all three models show an increase in usage over

time, with Model 1 showing the largest increase (5%).

Variables of Interest

Long term care insurance was a binary time varying variable where yes = 1 and no = 0.

Any responses of don’t know were left as missing. While one could suggest that not knowing

could be the same as not having insurance when trying to predict usage of benefits, however, this

decision was made because it is possible an individual was receiving advice from others who do

know about their insurance and factored it into their recommendations.

Table 1

Descriptive Statistics for Dependent Variables Across Waves

2006

n = 7069

2008

n=5867

2010

n=11,084

2012

n=9806

2014

n=8362

Uses SSDI and/or SSI (Model 1) 11.50% 11.97% 13.33% 14.84% 16.47%

Uses SSI (Model 2) 3.11% 3.51% 4.11% 5.11% 5.36%

Uses SSDI (Model 3) 9.15% 9.49% 11.43% 12.02% 13.81%

Sources: Health and Retirement Study, RAND HRS 2006-2014 and Psychosocial Leave Behind 2008/10


Health insurance was a time varying binary variable where individuals who had insurance

due to their own employer, their spouse’s employer, or covered but don’t know whether by

employer or not were coded as 1. Those without were coded as 0.

Psychiatric diagnosis was time invariant and binary where if the respondent was asked

every wave if they have ever been told by a doctor that they had psychiatric problems they were

coded as 1 and if not they were 0. Missing and don’t know responses were left as missing

responses.

Big 5 personality traits were pulled from the psychosocial leave behind survey. Since only

50% of respondents were interviewed each wave (every other year), responses from 2008 and

2010 were combine to create a time invariant variable for each trait; extroversion, agreeableness,

conscientiousness, openness to experience, and neuroticism. Each scale was coded based on the

guidance of Smith et al. (2013) and ranged from 1 – 4, where 1 = not at all, 2 = a little, 3 = some,

and 4 = a lot. While some researchers have found personality traits to change over time due to

sudden life events and can be curvilinear in some cases, they also found a peak is reached between

age 40 and 60 (Specht, Egloff, & Schmukle, 2011). A longer standing belief is that personality

traits are relatively stable over time (Roberts, Wood, & Caspi, 2008). Given this time frame only

covers one decade and the well-grounded research that personalities are relatively stable, it was

decided to create this variable as time invariant.

Extroversion included sub-dimensions of outgoing, friendly, lively, active, and talkative.

Agreeableness included sub-dimensions of helpful, warm, caring, softhearted, and sympathetic.

Conscientiousness included the sub-dimensions of organized, responsible, hardworking, careless

(reverse coded), and thorough. Openness to experience included the sub-dimensions of creative,

imaginative, intelligent, curious, broadminded, sophisticated, and adventurous. Neuroticism

included the sub-dimensions of moody, worrying, nervous, and calm (reverse code).

A scale for each trait was created by calculating the means of the sub-dimension within a

given trait for each respondent, shown in Table 2. For each scale, the trait was set to missing if

more than half of the items had missing values in within each sub-dimension. All sub-dimensions

except for careless and calm were on a scale of 1 – 4, where 1 = not at all, 2 = a little, 3 = some,

and 4 = a lot. Careless and calm were reverse coded.

Employer accommodations represented two different variables explored. The first was

whether an employer provided accommodations at the onset of disability limiting one’s ability to

work and if a current employer is providing accommodations. While the variables were not

ultimately utilized in the models several efforts were made to incorporate them into the analysis.

Control Variables

Income was used as a time varying continuous control variable. In order to provide a

normal distribution, the income was logged. To avoid the issue of logging a zero, $1 was added to

the income variable. Also time varying and continuous was net worth. Net worth was transformed

using the inverse hyperbolic sine to allow for utilization of negative net worth responses (see

Pence, 2006).


Labor force had the reference category of disabled and the other categories included full

time, part time, unemployed, part time retired, retired, and not in labor force. Another categorical

variable was marital status. Married was the reference group and compared to partnered, divorced

or separated, widowed, and never married. Divorced and separated were combined because they

are similar in that the single economic unit concept no longer fits and a built in physical/social

support system is not present.

Self-reported health was a continuous time variant variable and reverse coded to be

ordered poor (1) to excellent (5). Gender was included as a time invariant binary control variable

where male = 1 and female = 0. A categorical variable of race included White as the reference

group compared to Black and other. Education in years was included as a time invariant

continuous variable. Census region was in categories of Northeast, Midwest, South, and West with

South being the reference group. Due to the small number of other responses and not wanting to

exclude them, individuals with other as a response were combined with West.

Empirical Models

With the longitudinal nature of the data available, need for some time invariant variables to

operationalize the theory, and a desire to determine the main effect of both time varying and time

invariant characteristics, random effects (or random slopes) logistic regression models were

estimated. Three random effects logit models were calculated as follows:

log (Ρ𝑖𝑡

1 − Ρ𝑖𝑡) = μ𝑡 + βx𝑖𝑡 + z𝑖 + α𝑖

Where Ρit is the probability householder i exhibits a characteristic in wave t, xit is a vector of time-

varying household characteristics, zi is a vector of time-invariant household characteristics, and αi

is a random variable representing all differences between individuals that are stable over time and

not accounted for by zi.

Findings/Results

The sample across the decade had a median age ranging from 56 to 58, were mostly

married (60.6-70.9%), White (76.6%), female (61.51%), employed full time (45.54-48.82%) or

retired (18.67-26.07%), had a median of 13 years of education, and almost a quarter (24.35%) had

been told by a doctor that they suffered from a psychiatric diagnosis at one time in their life. As

shown in Table 2, individuals having long term care insurance ranged over time from 8% to over

10% and employer provided health insurance coverage varied over time from approximately 55%

– 69% of respondents.

Model 1, shown in Table 3, predicted use of SSI and/or SSDI in individuals aged 50-65

and presented significant findings. With a rho of .9147, the model suggested approximately 92%

of variation in outcome was due to individual changes over time. Individuals with health insurance

provided by a spouse’s or one’s own employer had approximately 93% lower odds of using SSI or

SSDI than individuals who did not. Those with a psychiatric diagnosis from anytime in their

lifetime had over 17 times the odds of using SSI and/or SSDI than someone without a diagnosis.


An individual with a unit increase on the Big 5 Agreeableness scale (1-4) had 2.7 times the odds

of using SSI and/or SSDI. However when separated into Model 2 (Table 4) and Model 3 (Table 5)

it was no longer significant. With a unit increase in the Big 5 Conscientiousness scale (1-4) an

individual had approximately 82% lower odds of using SSI or SSDI.

Marital status played a very important role in use of benefits; divorced/separated and

partnered had almost 5 times higher odds of using SSI and/or SSDI than married individuals and

never married had over 5 times the odds over married individuals. For each additional year of

education achieved, an individual had approximately 17% lower odds of using SSI and/or SSDI.

Each wave when compared to the first wave had 1 to 4 times the odds of using SSI and/or SSDI.

Model 2, shown in Table 4, predicts usage of SSI. The only difference in the setup from

Model 1 to Model 2 is the independent variable. Model 2 had a rho of .7819 suggesting

approximately 78% of variation in the outcome was due to individual changes over time in this

model. Health insurance and psychiatric diagnosis continued to be significant predictors. Unique

to Model 2 (Table 4), individuals with long term care insurance had approximately 67% lower

odds of using SSI than individuals without. While this was not significant in Model 1 with SSI and

SSDI combined it is significant when predicting SSI usage alone. As mentioned previously,

agreeableness was no longer significant in this separate SSI model when it was in the combined

SSI and/or SSDI model. Also different was only wave 4 when compared to wave 1 was significant

and showed an individual had almost 1.7 times the odds of using SSI in wave 4 when compared to

wave 1. Models 1 and 3 showed more significant and greater magnitude of prediction across

waves.

The last model, Model 3, predicts use of SSDI. This model, shown in Table 5, had a slight

variation of Models 1 and 2 because it did not include the use of long term care insurance. Since

SSDI does not provide long term care benefits it was removed from the model as a choice

substitution. The results were consistent with Model 1 except for agreeableness trait was no longer

significant, just as it was not in Model 2. Also unique was only waves 3 through 5 when

compared to the first wave significantly predicted the use of SSDI. An individual had over 2 times

the odds of using SSDI in each wave 3-5 than in wave 1. Model 3 had a rho of .8731 suggesting

approximately 87% of variation in the outcome was due to individual changes over time in this

model.

As hypothesized, individuals with long term care insurance were less likely to use SSI than

those without long term care insurance. While Model 1 with both SSI and/or SSDI was not

significant, when predicting only SSI in Model 2, individuals with long term care insurance had

approximately 67% lower odds of using SSI. Also, having employer provided health insurance

either from their own employer or spouse’s employer significantly reduced the odds of using SSI

and/or SSDI in all three models and were as hypothesized.

While individuals use of SSI and SSDI did vary across personality traits, only

agreeableness and conscientiousness were found to be significant. The directionality of

agreeableness was opposite of that hypothesized and was only found to be significant in Model 1

where SSI and/SSDI were predicted. However, and increase in the conscientiousness scale

decreased the odds of using SSI and/or SSDI in all three models.


Since the dependent variable of SSI usage was not able to take into account any of the

individuals who responded they were using SSI/SSDI but did not know which, the restricted data

were used by merging it with the public data. The SSI usage variable was recoded to identify

anyone who did not know whether they were using SSI/SSDI and were included in the SSI

restricted sample as user of SSI. This grew the sample from 7210 to 7244. All variables of

significance remained, directionality was consistent, and magnitude along with rho were very

similar. Since there were little to no differences between the merged restricted and public data

sets, the public information is reported.


Table 2

Descriptive Statistics for Independent Variables Across Waves, Proportion or Median

Proportion or Median2006

n = 7069

2008

n=5867

2010

n=11,084

2012

n=9806

2014

n=8362

Long term care insurance 9.77% 10.23% 8.58% 8.03% 8.13%

Health insurance (R or Sp employer) 68.80% 66.62% 58.91% 55.67% 55.17%

Psychiatric diagnosis

Personality traits (scale 1-4)

Extroversion

Agreeableness

Conscientiousness

Openness

Neuroticism

HH income (logged) 55,500$ 59,600 50,000 50,000 50,000

HH net worth (invhypsine) 190,300$ 174,000 80,000 76,400 85,000

Labor force

Disabled 4.58% 4.45% 4.63% 4.56% 4.02%

Full time 48.78% 48.82% 47.97% 47.22% 45.54%

Part time 9.49% 9.17% 10.67% 10.44% 10.11%

Unemployed 2.40% 2.86% 6.90% 5.82% 4.09%

Part time retired 6.10% 6.00% 4.23% 4.69% 5.12%

Retired 21.09% 21.66% 18.67% 22.25% 26.07%

Not in labor force 7.55% 7.04% 6.93% 5.02% 5.06%

Marital status (married)

Married 70.90% 69.46% 63.05% 61.84% 60.60%

Partnered 5.53% 5.80% 7.62% 7.82% 8.31%

Divorced/separated 14.09% 14.47% 16.99% 17.50% 18.19%

Widowed 5.96% 6.53% 5.21% 5.47% 5.54%

Never married 3.49% 3.73% 7.09% 7.36% 7.33%

Health self rep (scale 1-5) 3.00 3.00 3.00 3.00 3.00

Male

Race

White

Black

Other

Education (years)

Census region

South 39.71% 39.99% 41.26% 40.88% 40.81%

North East 14.67% 14.06% 15.37% 15.29% 84.51%

Midwest 25.13% 25.05% 19.70% 19.69% 19.29%

West (and other) 20.49% 20.87% 23.67% 24.14% 24.41%

Age 57 58 56 57 58

Note: Unweighted. Personality traits sampled half in 2008 and half in 2010.


13.0 - time invariant


38.49% - time invariant










Table 3

Uses SSI or SSDI OR SE B P>z

Long term care insurance 0.8992 0.2797 0.733 0.4888 1.6544

Health insurance (R or Sp employer) 0.0652 *** 0.0155 <.001 0.0410 0.1039

Psychiatric diagnosis 17.5723 *** 5.4807 <.001 9.5356 32.3826

Personality traits

Extroversion 0.6153 0.1781 0.093 0.3489 1.0850

Agreeableness 2.6973 ** 0.9088 0.003 1.3936 5.2207

Conscientiousness 0.1835 *** 0.0584 <.001 0.0983 0.3425

Openness 1.0747 0.3111 0.804 0.6094 1.8953

Neuroticism 1.1218 0.2465 0.601 0.7292 1.7258

HH income (logged) 1.0821 0.0537 0.112 0.9817 1.1927

HH net worth (invhypsine) 0.9586 ** 0.0120 0.001 0.9354 0.9823

Labor force (disabled)

Full time 0.0001 *** 0.0000 <.001 0.0000 0.0002

Part time 0.0004 *** 0.0002 <.001 0.0001 0.0014

Unemployed 0.0017 *** 0.0010 <.001 0.0006 0.0051

Part time retired 0.0068 *** 0.0036 <.001 0.0024 0.0193

Retired 0.6538 0.1796 0.122 0.3816 1.1202

Not in labor force 0.0141 *** 0.0071 <.001 0.0052 0.0379


Partnered 4.0279 ** 1.7207 0.001 1.7436 9.3046

Divorced/separated 6.2319 *** 2.0462 <.001 3.2744 11.8604

Widowed 1.6182 0.6238 0.212 0.7602 3.4447

Never married 3.7199 * 1.9176 0.011 1.3544 10.2168

Health self rep 0.2245 *** 0.0259 <.001 0.1791 0.2814

Male 2.4647 ** 0.6970 0.001 1.4160 4.2901

Race (white)

Black 8.7976 *** 3.0496 <.001 4.4597 17.3550

Other 0.7464 0.3427 0.524 0.3035 1.8354

Education (years) 0.8371 *** 0.0378 <.001 0.7663 0.9144

Census region (South)

North East 2.9973 ** 1.0863 0.002 1.4731 6.0985

Midwest 1.4281 0.4483 0.256 0.7718 2.6422

West 0.8048 0.2683 0.515 0.4187 1.5470

Wave

2 1.5868 * 0.3078 0.017 1.0849 2.3208

3 2.2244 ** 0.5222 0.001 1.4040 3.5241

4 3.0537 *** 0.7862 <.001 1.8437 5.0580

5 4.2322 *** 1.1178 <.001 2.5220 7.1021

_cons 49.5923 * 76.0919 0.011 2.4512 1003.3540

sigma_u 5.9388 0.3201 5.3433 6.6006

rho 0.9147 0.0084 0.8967 0.9298

*p < .05. **p < .01. ***p < .001.

Summary of Random Effects Logistic Regression Analysis for Variables Predicting Social Security

Disability (SSDI) Participation or Supplemental Security Income (SSI) Either Through Pending

Application or Receipt by Individuals Under the Age of 65 (n = 7210)

Note: Standard errors are robust. Health insurance coverage include both sources of respondent and

spouse employer provided. Personality traits were measured as time invariant (measured in 2008 or

2010) and formed using scales for the "Big 5" Personality Traits. Income was transformed by logging and

net worth was transformed using inverse hyperbolic sine.

[95% Conf. Interval]



Table 4

Uses SSI Disability OR SE B P>z

Long term care ins 0.3260 * 0.1432 0.011 0.1378 0.7713



Personality traits

Extroversion 0.8784 0.2165 0.599 0.5419 1.4239

Agreeableness 1.0829 0.3152 0.784 0.6120 1.9159

Conscientiousness 0.3767 *** 0.0977 <.001 0.2266 0.6262

Openness 1.5312 0.3795 0.086 0.9420 2.4890

Neuroticism 1.0781 0.2026 0.689 0.7460 1.5583

HH income (logged) 1.0693 0.0429 0.095 0.9883 1.1568

HH net worth (invhypsine) 0.9534 *** 0.0104 <.001 0.9333 0.9739


Full time 0.0023 *** 0.0014 <.001 0.0007 0.0075

Part time 0.0030 *** 0.0022 <.001 0.0007 0.0127

Unemployed 0.0163 *** 0.0094 <.001 0.0053 0.0503


Retired 0.6338 0.1328 0.03 0.4202 0.9557



Partnered 4.6561 *** 1.6135 <.001 2.3608 9.1830

Divorced/separated 4.8116 *** 1.2487 <.001 2.8933 8.0019

Widowed 1.1995 0.3903 0.576 0.6340 2.2697

Never married 5.4627 *** 1.9306 <.001 2.7326 10.9203

Health self rep 0.5217 *** 0.0542 <.001 0.4256 0.6395

Male 0.4841 ** 0.1207 0.004 0.2969 0.7893

Race (white)

Black 3.7579 *** 1.0154 <.001 2.2128 6.3818

Other 1.8044 0.6494 0.101 0.8912 3.6534

Education (years) 0.8338 *** 0.0317 <.001 0.7740 0.8983


North East 2.1547 * 0.6910 0.017 1.1493 4.0398

Midwest 1.6979 0.4972 0.071 0.9564 3.0142

West 2.4749 ** 0.7521 0.003 1.3642 4.4900

Wave

2 1.3344 0.2858 0.178 0.8770 2.0303

3 1.2867 0.3029 0.284 0.8111 2.0409

4 1.6614 * 0.4184 0.044 1.0141 2.7217

5 1.6616 0.4350 0.052 0.9947 2.7758

_cons 0.5212 0.6581 0.606 0.0439 6.1915

sigma_u 3.4341 0.1973 3.0684 3.8434

rho 0.7819 0.0196 0.7411 0.8179

*p < .05. **p < .01. ***p < .001.



Summary of Random Effects Logistic Regression Analysis for Variables Predicting Supplemental

Security Income (SSI) Participation Either Through Pending Application or Receipt by Individuals

Under the Age of 65 (n = 7210)

Note: Standard errors are robust. Personality traits were measured as time invariant (measured in 2008

or 2010) and formed using scales for the "Big 5" Personality Traits. Income was transformed by logging

and net worth was transformed using inverse hyperbolic sine.


Table 5

Uses SSDI OR SE B P>z



Personality traits

Extroversion 0.7591 0.1865 0.262 0.4689 1.2288

Agreeableness 1.7943 0.4806 0.029 1.0615 3.0331

Conscientiousness 0.4442 ** 0.1115 0.001 0.2716 0.7266

Openness 0.9458 0.2206 0.811 0.5988 1.4938

Neuroticism 0.9595 0.1691 0.814 0.6793 1.3552

HH income (logged) 1.0324 0.0412 0.424 0.9548 1.1164

HH net worth (invhypsine) 0.9686 ** 0.0096 0.001 0.9500 0.9875


Full time 0.0005 *** 0.0003 <.001 0.0002 0.0014

Part time 0.0032 *** 0.0016 <.001 0.0012 0.0087

Unemployed 0.0132 *** 0.0055 <.001 0.0058 0.0300


Retired 0.9703 0.2097 0.889 0.6352 1.4822



Partnered 1.7526 0.6070 0.105 0.8889 3.4554

Divorced/separated 1.9592 ** 0.4926 0.007 1.1968 3.2070

Widowed 1.6498 0.4616 0.074 0.9534 2.8548

Never married 1.4296 0.5743 0.374 0.6506 3.1415

Health self rep 0.2948 *** 0.0253 <.001 0.2491 0.3488

Male 3.0458 *** 0.6870 <.001 1.9575 4.7391

Race (white)

Black 4.1131 *** 1.1116 <.001 2.4217 6.9858

Other 0.6810 0.2481 0.292 0.3335 1.3908

Education (years) 0.9418 0.0331 0.088 0.8791 1.0091


North East 1.4322 0.4295 0.231 0.7957 2.5778

Midwest 1.2014 0.3058 0.471 0.7295 1.9784

West 0.5595 * 0.1528 0.034 0.3276 0.9557

Wave

2 1.2852 0.2199 0.143 0.9190 1.7971

3 2.1804 *** 0.4132 <.001 1.5039 3.1613

4 2.1543 *** 0.4485 <.001 1.4325 3.2398

5 2.8649 *** 0.6037 <.001 1.8955 4.3301

_cons 2.1483 2.6382 0.534 0.1935 23.8459

sigma_u 4.757931 0.218385 4.348593 5.2058

rho 0.873114 0.01017 0.8518086 0.8917458

*p < .05. **p < .01. ***p < .001.



Summary of Random Effects Logistic Regression Analysis for Variables Predicting Social Security

Disability (SSDI) Participation Either Through Pending Application or Receipt by Individuals Under

the Age of 65 (n = 7221)

Note: Standard errors are robust. Health insurance coverage include both sources of respondent and

spouse employer provided. Personality traits were measured as time invariant (measured in 2008 or

2010) and formed using scales for the "Big 5" Personality Traits. Income was transformed by logging

and net worth was transformed using inverse hyperbolic sine.


Discussion/Implications

Employer provided health insurance can be valuable in reducing the odds of individuals

using SSI and/or SSDI. Long term care insurance can be another useful tool in reducing the odds

of individuals using SSI which comes directly from the US Treasury General Funds. All citizens

have a vested interest in reducing the expenses of welfare but not at the expense of the individuals

receiving the benefits. If benefits can be self-funded, employer funded or incentivized then both

needs are met. All labor force categories except for retirement have significantly lower odds of

using SSI and/or SSDI than individuals reporting as disabled for the labor force questions. This is

consistent with expectations given the necessary requirements to qualify for the programs.

When modeling the use of SSDI and/or SSI, use of SSI only, use of SSDI only, further

evidence of the mental health crisis is provided. There is a great need to take measures to increase

the mental wellness of America and support of those suffering from mental illness. Additionally,

the results provide further evidence of the value of education. Women have lower odds of using

the benefits. Married individuals have the lowest odds and that is likely attributed to the married

couple being considered one single economic unit and having the support of each other. There is

also a lower expected need for long term care services since there is presumed to be a spouse

available to assist.

As education increases, the odds of being on SSI and/or SSDI are significantly lower.

Future research includes examining the length of individuals receiving benefits. For the same

reasons as having higher education resulting in lower odds of using benefits, it could be expected

that higher education would result in shorter terms of receiving benefits.

This study provides a glimpse into personality traits and disability benefit usage. Given the

previous findings of the personality trait of agreeableness being positively associated with

adaption and recovery of life satisfaction after an acquired disability (Boyce & Wood, 2011) along

with the positive relationship found here with use of SSI and/or SSDI and insignificant results in

Models 2 (SSI only) and 3 (SSDI only), further research is needed. Further investigation into the

traits and ways to leverage the knowledge of those traits in individuals can help inform future

policy measures.

After completing phase 1 analyzing the public data, the combined benefits logit resulted in

the long term care insurance independent variable not being significant and the agreeableness

scale as significant. Long term care was not included in the SSDI analysis since SSDI does not

provide that coverage, but it was significant in the SSI only coverage. When broken up into the

SSDI model and the SSI model, agreeableness was no longer significant. The publicly available

data reported a substantial number of individuals who indicated they were applying for or

receiving some benefit but did not know which benefit. By incorporating in the more detailed

restricted data 20 more respondents were moved from “did not know” to “applying for or

receiving SSI”. It was hypothesized that the “did not know” respondents made an impact on the

agreeableness scale. Even after adding in additional respondents as using SSI, agreeableness was

still not significant. Further analysis of “did not know” group could prove to be beneficial, but it

could represent lack of knowledge therefore lack of calculated decisions in deciding to use

benefits or not. If lack of knowledge is the case, then educational interventions could be beneficial.


Employer accommodations at the onset of disability diagnosis and accommodations in a

current position were removed from the model. Many different models were explored to allow for

the inclusion of current and prior employer accommodations, but the regressions would not

function properly resulting in insignificance across the model, lack of fit, and multicollinearity

issues. The low frequency appeared to cause the greatest problems in the models. The low

frequency was attributed in large part to the sample being individuals nearing retirement and the

fact that respondents had to be working to report receiving employer accommodations. Very few

individuals reported currently receiving employer accommodations or having received them upon

diagnoses. Prior research found personality traits were highly predictive of employer

accommodations (Hill, Maestas, & Mullen, 2016) which could explain multicollinearity issues

when employer accommodations were included.

An investigation was conducted into respondents who were using SSI/SSDI and then not in

the following wave, however, the frequencies were too low to conduct a meaningful analysis. For

example, there were only 180 respondents using SSI in wave 4 who were not using SSI in wave 5.

Of the 180, only 1 was receiving current employer accommodations. Further exploration should be

considered in alternate datasets not limited to individuals over the age of 50. Future direction of

research includes examining individuals who were once receiving SSI and/or SSDI and no longer

receiving the benefits and the relationship of work accommodation. Examining receiving

SSI/SSDI rather than using it which includes applying for and/or receiving will be a different

approach considered in the future.

Based on the findings in this study, policy implications are suggested. Evidence suggests

continuing tax benefits for employer provided health plans could assist in reducing usage of SSI

and/or SSDI. Congress could consider expanding the tax benefits to those who provide family and

spousal coverage. Given the specific results related to SSI, which is funded by taxpayer dollars,

considering tax incentives for private long term care insurance and possibly even as an employer

provided or subsidized benefit is advisable.

Given the magnitude of increased odds for individuals with a mental health diagnosis,

considering incentivizing expanded health insurance to include more mental health coverage could

be worthwhile. Lastly, further evidence is provided to continue to provide higher education tax

incentives as increased education lowers the odds of using SSI and/SSDI.

As Mitra suggests in the capability approach, the choices an individual makes can result in

actual function or lack thereof. Those who choose to purchase private long term care insurance are

less likely to lack function when defined as using SSI and/or SSDI. If positive choices can be

incentivized, with education and financial sophistication, disabled individuals who are insurable

for long term care insurance can rely on private insurance rather than spending down assets to

qualify to Medicaid. By adding incentives for Americans to purchase private long-term care

insurance action can lead to lower reliance on Medicaid and fewer individuals utilizing asset spend

down techniques.

While self-reported use of SSI and/SSDI is the only option provided in these data, along

with only a limited number of additional details in the restricted data, actual use of SSI and/or

SSDI would be helpful. Given the variation seen when comparing models, the individuals who


report using either SSI and/or SSDI but do not know which one are a unique group. If their

choices do impact their use of benefits, then were those choices guided by a professional? If not,

then they were ultimately uninformed choices but choices none the less. By using linked data from

the Social Security Administration, further insight can be gained into this population of

individuals. Another limitation to this study was lack of information on private disability

insurance. Exploring other sources which include information on disability insurance could be

information.

In support of the goal of ARDRAW related to employer supports, this research suggests

employers might play a key role in reducing the odds of their employees or their spouses using

SSI and/or SSDI. Health insurance provided by employers of either the respondent or spouse

decreased the odds of using SSI and/or SSDI in all three regressions. While this research does not

suggest a causal relationship between having employer provided health insurance and the use of

SSDI and/or SSI, it does demonstrate a decrease in the odds of usage. The same goes for a

reduction in odds of using SSI when an individual has long term care insurance rather than a

causal relationship. Long term care benefits could be offered by employers to make acquisition

more affordable for employees since income could be a limiting factor. Similar to prior research

findings, having a substitute benefit decreased the odds of benefits usage (Brown, Goda, &

McGarry, 2012). Employer provided substitutes appear to be worth exploring and possibly

incentivizing.


References

2017 Annual Report of the SSI Program (2018). Federal payments under SSI. Retrieved from

https://www.ssa.gov/OACT/ssir/SSI17/IV_C_Payments.html#567416

Altman, B. M. (2001). Definitions, Models, Classifications, Schemes, and Applications.

Handbook of Disability Studies, 97–122. doi:10.4135/9781412976251.n4

Andersen, M. F., Nielsen, K. M., & Brinkmann, S. (2012). Meta-synthesis of qualitative research

on return to work among employees with common mental disorders. Scandinavian Journal

of Work, Environment & Health, 38(2), 93–104. doi:10.5271/sjweh.3257

Bernell, S. L. (2003). Theoretical and Applied Issues in Defining Disability in Labor Market

Research. Journal of Disability Policy Studies, 14(1), 36-45.

doi:10.1177/10442073030140010501

Blustein, D. L. (2008). The role of work in psychological health and well-being: A conceptual,

historical, and public policy perspective. American Psychologist, 63(4), 228.

doi:10.1037/0003-066x.63.4.228

Boyce, C. J., & Wood, A. M. (2011). Personality prior to disability determines adaptation:

Agreeable individuals recover lost life satisfaction faster and more completely.

Psychological Science, 22(11), 1397-1402. doi:10.1177/0956797611421790

Brown, J. R., & Finkelstein, A. (2008). The interaction of public and private insurance: Medicaid

and the long-term care insurance market. American Economic Review, 98(3), 1083-1102.

doi:10.1257/aer.98.3.1083

Brown, J. R., Goda, G. S., & McGarry, K. (2012). Long-term care insurance demand limited by

beliefs about needs, concerns about insurers, and care available from family. Health

Affairs, 31(6), 1294-1302. doi:10.1377/hlthaff.2011.1307

Burkhauser, R. V., Butler, J. S., Kim, Y. W., & Weathers, R. R. (1999). The importance of

accommodation on the timing of disability insurance applications: Results from the survey

of disability and work and the health and retirement study. Journal of Human Resources,

34(3), 589-611. doi:10.2307/146381

Burkhauser, R. V., & Stapleton, D. C. (2004). The decline in the employment rate for people with

disabilities: Bad data, bad health, or bad policy?. Journal of Vocational Rehabilitation,

20(3), 185-201.

Caspi, A., Roberts, B. W., & Shiner, R. L. (2005). Personality development: Stability and change.

Annual Review of Psychology, 56, 453-484.

doi:10.1146/annurev.psych.55.090902.141913


Corrigan, P. W., & Watson, A. C. (2004). At issue: Stop the stigma: Call mental illness a brain

disease. Schizophrenia Bulletin, 30(3), 477. doi:10.1093/oxfordjournals.schbul.a007095

Erickson, W., Lee, C., & von Schrader, S. (2018). 2016 Disability Status Report: United States.

Ithaca, NY: Cornell University Yang-Tan Institute on Employment and Disability (YTI).

Hill, M. J., Maestas, N., & Mullen, K. J. (2016). Employer accommodation and labor supply of

disabled workers. Labour Economics, 41, 291-303. doi:10.1016/j.labeco.2016.05.013

Kapteyn, A., Michaud, P. C., Smith, J. P., & Van Soest, A. (2006). Effects of attrition and non-

response in the Health and Retirement Study. RAND Working Paper WR-407.

Laughlin, C., & Gotten, P. D. (2008). Efficacy of a pre-retirement planning intervention for ageing

individuals with mental retardation. Journal of Intellectual Disability Research, 38(3),

317–328. doi:10.1111/j.1365-2788.1994.tb00398.x

Low, H., & Pistaferri, L. (2010). Disability risk, disability insurance and life cycle behavior (No.

w15962). National Bureau of Economic Research.

Maestas, N., Mullen, K. J., & Strand, A. (2013). Does disability insurance receipt discourage

work? Using examiner assignment to estimate causal effects of SSDI receipt. The

American Economic Review, 103(5), 1797-1829. doi:10.1257/aer.103.5.1797

McAlpine, D. D., & Warner, L. (2000). Barriers to employment among persons with mental

illness: A review of the literature. Center for Research on the Organization and

Financing of Care for the Severely Mentally Ill Institute for Health, Health Care Policy

and Aging Research Rutgers, the State University: New Brunswick, NJ.

Mitra, S. (2006). The capability approach and disability. Journal of Disability Policy Studies,

16(4), 236-247. doi:10.1177/10442073060160040501

Mroczek, D. K., Spiro, A., & Griffin, P. W. (2006). Personality and aging. In J. E. Birren & K. W.

Schaie (Eds.), Handbook of the Psychology of Aging (6th ed., pp. 363-377). Burlington,

MA: Elsevier Press. doi:10.1016/b978-012101264-9/50019-7

Ozer, D. J., & Benet-Martinez, V. (2006). Personality and the prediction of consequential

outcomes. Annual Review of Psychology, 57, 401-421.

doi:10.1146/annurev.psych.57.102904.190127

Pauly, M. V. (1990). The rational nonpurchase of long-term-care insurance. Journal of Political

Economy, 98(1), 153-168. doi:10.1086/261673

Pence, K. M. (2006). The Role of Wealth Transformations: An Application to Estimating the

Effect of Tax Incentives on Saving. Contributions in Economic Analysis & Policy, 5(1).

doi:10.2202/1538-0645.1430


Roberts, B., Kuncel, N.R., Shiner, R., Caspi, A., & Goldberg, L.R. (2007). The power of

personality: The comparative validity of personality traits, socioeconomic status, and

cognitive ability for predicting important life outcomes. Perspectives on Psychological

Science, 2, 313-345. doi:10.1111/j.1745-6916.2007.00047.x

Roberts, B. W., Wood, D., & Caspi, A. (2008). The development of personality traits in adulthood.

In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of Personality: Theory

and Research (3rd ed., pp. 375-398). New York: The Guilford Press.

Schur, L. A. (2003). Barriers or opportunities? The causes of contingent and part‐time work

among people with disabilities. Industrial Relations: A Journal of Economy and Society,

42(4), 589-622.

Sloan, F. A., & Norton, E. C. (1997). Adverse Selection, Bequests, Crowding Out, and Private

Demand for Insurance: Evidence from the Long-Term Care Insurance Market. Journal of

Risk and Uncertainty, 15(3), 201-219. doi:10.1016/s0167-6687(98)80095-4

Smith, J., Fisher, G., Ryan, L., Clarke, P., House, J., & Weir, D. (2013). Psychosocial and

Lifestyle Questionnaire, 2006 – 2010. Retrieved from

http://hrsonline.isr.umich.edu/sitedocs/userg/HRS2006-2010SAQdoc.pdf

Social Security and Medicare Boards of Trustees (2018). A summary of the 2018 annual reports.

Retrieved from https://www.ssa.gov/oact/trsum/

Solovieva, T. I., Dowler, D. L., & Walls, R. T. (2011). Employer benefits from making workplace

accommodations. Disability and Health Journal, 4(1), 39-45.

doi:10.1016/j.dhjo.2010.03.001

Specht, J., Egloff, B., & Schmukle, S. C. (2011). Stability and change of personality across the life

course: The impact of age and major life events on mean-level and rank-order stability of

the Big Five. Journal of Personality and Social Psychology, 101(4), 862.

doi:10.1037/a0024950

Stein, M. A. (2003). The law and economics of disability accommodations. Duke Law Journal, 53,

79.

Unger, D. D. (2002). Employers' attitudes toward persons with disabilities in the workforce: myths

or realities? Focus on Autism and Other Developmental Disabilities, 17(1), 2-10.

doi:10.5271/sjweh.3257


Appendix

Figure 1: The Capability Approach (Mitra, 2006)

running head: ssi and ssdi usage among americans nearing ......disability theories, and disability...

Documents